Skip to content

NEWS.md

Latest commit

 

History

History
61 lines (58 loc) · 3.74 KB

NEWS.md

File metadata and controls

61 lines (58 loc) · 3.74 KB
Raw

mutt 0.1.0.9000

  • Functional publication now writes a compact functional-core.zip plus separately published EC, KO, and MetaCyc contribution assets. Oversized contribution tables are sharded below the configured per-file limit, while mutt() transparently reads legacy loose files and new bundle descriptors.
  • Added functional = "REVALIDATE" to reconstruct validated PICRUSt2 results from retained normal or .failed/raw output without launching PICRUSt2. Pathway-only sample omissions are zero-filled and reported as generated_with_warning; EC/KO disagreement and unexpected pathway samples remain failures.
  • The HPC study runner now accepts revalidate and exits with status 11 when any functional manifest branch has status = "failed", preventing failed studies from receiving completion checkpoints.
  • Functional inference now routes only explicitly identified amplicon branches to PICRUSt2 and FAPROTAX; shotgun, metagenomic, and unknown-modality branches are retained in the manifest as skipped.
  • PICRUSt2 EC and KO outputs must agree exactly. Pathway outputs may omit EC/KO samples, which are restored as zero-prediction rows with explicit reconciliation QC; unexpected pathway samples remain validation failures.
  • DADA2 PICRUSt2 inputs now merge exact reverse-complement ASV duplicates before inference while preserving every sample's read total and rebuilding aligned consensus taxonomy for the merged features.
  • Added safe, explicit one-file study-data archiving with Git LFS verification.
  • Study TSV and RDS readers now support either an uncompressed file or its one-file .zip archive, preserving local rebuild workflows and archived Git checkouts.
  • Functional caches can be restored from validated functional.zip archives; cached PICRUSt2 output paths are rebased to the current checkout.
  • Successful PICRUSt2 caches discard unneeded orientation and intermediate alignment directories after parsing while retaining predictions, stratified contribution files, mappings, taxonomy, QC, logs, and provenance.
  • Main-branch container builds are published to GitHub Container Registry for conversion and execution with Apptainer on HPC systems.
  • Restructured MUTT as an installable package with mutt() as its sole public API, bundled parsers, local/remote study resolution, structured audit records, and checksum-verified atomic caching.
  • Added all-study data-release readiness, build, local verification, and remote verification tooling. Production remote flags remain disabled until assets are published and verified.
  • Added optional cached PICRUSt2 and FAPROTAX inference through mutt(functional = FALSE), with TRUE and "REBUILD" modes.
  • Functional files are stored under each study's functional/ directory and returned in the study's function entry with branch-level provenance.
  • PICRUSt2 inputs now undergo per-ASV strand selection against its bacterial and archaeal reference HMMs, with alignment fractions and decisions retained.
  • PICRUSt2 now always requests stratified EC, KO, and MetaCyc output. Large ASV-contribution tables remain file-backed and are accessed through as.data.frame() on a returned PICRUSt2 branch.
  • Bundled FAPROTAX 1.2.12 files are discovered without directory arguments.
  • FAPROTAX now prefers matched count and taxonomy branches and falls back to matched proportion and taxonomy branches when counts are unavailable.
  • Preserved valid sequence columns when taxonomy row names are ASV identifiers.
  • Added a compatibility audit for the 33 study outputs present in the saved pre-restructure validation artifact. The audit records that this baseline used sample alignment and keeps it separate from the default-mode parser inventory.

mutt 0.1.0

  • Initial CRAN submission.